Process a lot of information quickly

• distributions, relationships, outliers

• summary statistics, correlations, e.g. do not tell the whole story

\(~\)

Different philosophies

• base R plots

  • one type of plot (e.g. scatter, box, histogram)

  • plot arguments for appearance (e.g. labels, colors, sizes, etc.)

  • add things later (e.g. points, lines) with low-level plot functions

• ggplot

  • add layers to a plot

  • change appearance of the plot in any of these layers

    • e.g. add points, lines, change labels, shapes, orientation, multiples, etc.

High-level plot functions create a plot (… are additional arguments)

plot(x, y = NULL, ...)

hist(x, ...)   

boxplot(formula, data, ...) 

barplot(formula, data, ...)     

Low-level plot functions add something to the plot

points(x, y)  

lines(x, y)   

abline(a, b, h, v)   

text(x, y, labels)      

legend(x, y, legend)      

par(mfrow = c(2, 3))
with(mtcars, plot(mpg))                    # or plot(mtcars$mpg)
with(mtcars, boxplot(mpg))
with(mtcars, plot(factor(gear)))
with(mtcars, plot(disp, mpg))
with(mtcars, boxplot(mpg ~ gear))
with(mtcars, barplot(xtabs(~ gear + cyl))) # no legend!

with(mtcars, plot(disp, mpg, col = c("red", "blue")[am + 1]))   # add colors 

## Warning in Ops.factor(am, 1): '+' not meaningful for factors

legend("topright", legend = c("automatic", "manual"),           # add legend
       col = c("red", "blue"), pch = 1, title =  "transmission")      

Layered plotting based on the book The Grammer of Graphics by Leland Wilkinsons.

• start at the bottom and work your way up

1. Data argument (what is the name of the data set)
2. Aesthetics (what goes to the x- and y-axes)
3. Geometrics (what is used to represent the data)

library(patchwork)                       # combine plots in one figure
ggplot(data = mtcars) + 
ggplot(mtcars, (aes(x = disp, y = mpg))) +
ggplot(mtcars, aes(disp, mpg)) + 
  geom_point()

It is possible to save plot object, and add layers later

gg <- ggplot(mtcars, aes(disp, mpg)) + 
  geom_point()
gg

1. Add a smooth regression line
2. Add colors for am (transmission) and loose the confidence band
3. Display separate linear regression lines for am

gg + geom_smooth() +
gg + geom_smooth(se = F) + 
  geom_point(aes(col = am)) +
gg + geom_smooth(aes(col = am), method = "lm", se = F) + 
  geom_point(aes(col = am))  + 
  plot_layout(guides = "collect", axes = "collect")  # collect labels/legends 

gg  + geom_smooth(aes(col = am), method = "lm", se = F) + 
  geom_point(aes(col = am)) +
  labs(x = "displacement", y = "miles per gallon", title = "Gas consumption") +
  theme_minimal()

• size
  • resize of geom according to value of a numeric variable
• color
  • color of geom on basis of a numeric/categorical variable
• fill
  • fill geom according to categories of a variable
• shape
  • change the shape of a geom according to categories of a variable
• alpha
  • change the opacity of a geom
• and others …

gg + geom_point(aes(col = am, size = hp)) +
  theme_minimal()

• geom_point

• geom_bar

• geom_line

• geom_smooth

• geom_histogram

• geom_boxplot

• geom_violin

• geom_density

• geom_bar

• and many more…

ggplot(mtcars, aes(mpg)) + geom_histogram(bins = 10) +
ggplot(mtcars, aes(mpg)) + geom_density() +
ggplot(mtcars, aes(mpg)) + geom_boxplot()

ggplot(mtcars) + geom_bar(aes(gear)) +
ggplot(mtcars) + geom_bar(aes(gear, fill = vs)) +
ggplot(mtcars) + geom_bar(aes(gear, fill = vs), position = "dodge") +
  plot_layout(guides = "collect", axes = "collect")

ggplot(mtcars, aes(mpg, am)) + geom_boxplot() +
ggplot(mtcars, aes(mpg, fill = am)) + geom_density(alpha = .3) +
ggplot(mtcars, aes(mpg, disp)) + geom_density_2d() 

Splitting the plot by levels of a categorical variable

gg + geom_smooth(method = "lm", se = F) +
  facet_grid(cols = vars(gear))

R Markdown has chunk options to resize plots. This is especially useful when more than one plot is displayed in a single plot array.

• fig.with = 7 is the default width of the array

• fig.height = 5 is the default height of the array

• fig.asp = .7 is the default height:width ratio

\(~\)

The defaults can be changed as follows:

```{r fig.asp = .4}

to reduce the height.